Blog - Konrad Kowalski (rootsher)
posts (31)
- CPU in Kubernetes from the kernel up2/12
cgroups and CPU: how Linux splits the processor between groups of processes
Group scheduling, cpu.weight and hierarchy - how the kernel divides CPU between groups of processes instead of individual tasks.
- CPU in Kubernetes from the kernel up1/12
How Linux really gives a process CPU
A process does not "have CPU". It is ready to run, and the scheduler decides when it actually gets processor time.
- series · 12 parts
CPU in Kubernetes from the kernel up
A CPU model built bottom-up: from runnable tasks and the scheduler, through cgroups, requests and limits, to throttling, contention, PSI and diagnostics.
- Frontend Rendering: from server-side pages to hybrid rendering17/17
Modern rendering is composition
Rendering stopped being a single stage. It is the placement of work across time, space and runtimes, and the framework is a consequence of those decisions.
- Frontend Rendering: from server-side pages to hybrid rendering16/17
Server Components: moving the execution boundary
The rendering boundary stops running through the page and starts running through the module graph. Some components never have a client-side implementation.
- Frontend Rendering: from server-side pages to hybrid rendering15/17
Resumability: what if we don't hydrate at all?
The server records enough information that the client never rebuilds the component tree. Code loads because execution became necessary.
- Frontend Rendering: from server-side pages to hybrid rendering14/17
Islands architecture: interactive fragments in a mostly static document
The starting point stops being the application. The page is a document, and only the fragments that truly need a client runtime get one.
- Frontend Rendering: from server-side pages to hybrid rendering13/17
Selective and progressive hydration: not everything deserves CPU now
Hydration stops being one startup task. Client CPU becomes a resource allocated by priority and by the user's real intent.
- Frontend Rendering: from server-side pages to hybrid rendering12/17
Edge rendering: compute moved closer to the user
Moving compute closer to the user only pays off when the data travels with it. Otherwise we shorten one hop and lengthen the rest.
- Frontend Rendering: from server-side pages to hybrid rendering11/17
Partial Prerendering: one page, several clocks
The route stops being the unit of decision. A static shell is produced earlier, dynamic fragments arrive at request time or through the stream.